You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
馃搳 Add Feature to Extract Record Holders Information from 10-K Filings聽#12
2600_yay brings attention to a tool that can be used to pull data from the SEC
They link to this repository, which provides a robust framework for extracting text from SEC documents. It can potentially be adjusted to extract data concerning record holders, which is crucial for financial analysis and investor relations.
Proposed Feature:
License: Confirm that the repository license allows for use with our cause.
Data Identification: Implement a method to accurately identify and extract the section of 10-K filings that lists the number of record holders. This often appears under the "Security Ownership of Certain Beneficial Owners and Management" section or similar headings.
Data Extraction: Develop a parsing function that can read through the identified section and extract the number of record holders. The function should handle variations in document formatting and text structures.
Output Specification: The extracted data should be output in a structured format (e.g., JSON, CSV) that specifies the company name, ticker symbol, and the number of record holders.
Integration: Ensure this new feature integrates seamlessly with the existing database and confirm the numbers are accurate. Strive for 95-99.9% accuracy.
Use Case: This feature will be particularly useful for analysts and investors looking to aggregate or compare shareholder data across different companies, providing a clearer picture of investor engagement and stock distribution.
Multiple phrase and sentence structures that refer to record holders would need to be identified and used.
Another challenge would be to identify issuers with multiple classes of stocks (Class A, Class B, etc.) and to be able to separate them accordingly. For now it may be useful to only extract record holder information of companies that have just one class of stock. Another issue can be created for a more refined extraction of multiple classes of stocks.
Ideally, this would be an automated script that runs daily and turns the data into a json file. Integrating that data into the database would be the final step; the .db file has CIK as the primary key and Ticker as the secondary key, so either of those columns should be used for merging purposes.
I did not look at the tool recommended by 2600, but is there a way to add general language, that AI can interpret, or do we need to open a few hundred 10Ks and see the language used and try to incorporate all that into the scraper? I am down to start building a term set if it will be helpful? maybe an alternate lane would be to scrape only item 5 in the 10Ks. IIRC , that is where the record holder is listed in nearly all of them that I have seen.
this is the list I was able to cobble together(6days found the last term)Holders of record, Registered holders, Share holders of record, Stockholders of record, Stockholders of record, Registered holders of record, Registered holders of record, Share owners of record, Registered shareholders, Common shareholders of record, edit, Registered stockholder, Record owner. if this needs to be itemized differently, LMK.
Activity
JamesAlfonse commentedon Sep 12, 2024
To add onto this, there are also extraction scripts available from @apes-on-parade here
Issuers are required to disclose the number of record holders each year in their 10K filing, but there is no strict requirement for the language used
Multiple phrase and sentence structures that refer to record holders would need to be identified and used.
Another challenge would be to identify issuers with multiple classes of stocks (Class A, Class B, etc.) and to be able to separate them accordingly. For now it may be useful to only extract record holder information of companies that have just one class of stock. Another issue can be created for a more refined extraction of multiple classes of stocks.
Ideally, this would be an automated script that runs daily and turns the data into a json file. Integrating that data into the database would be the final step; the .db file has CIK as the primary key and Ticker as the secondary key, so either of those columns should be used for merging purposes.
JFWooten4 commentedon Feb 5, 2025
Love the massive amount of collaboration hiding inbetween the lines on your work here James! Yay for help from 2600 on the Discord side 馃挏
bobmahalo commentedon Apr 27, 2025
I did not look at the tool recommended by 2600, but is there a way to add general language, that AI can interpret, or do we need to open a few hundred 10Ks and see the language used and try to incorporate all that into the scraper? I am down to start building a term set if it will be helpful? maybe an alternate lane would be to scrape only item 5 in the 10Ks. IIRC , that is where the record holder is listed in nearly all of them that I have seen.
JamesAlfonse commentedon Apr 28, 2025
A list of terms would be fantastic Bob. Any type of common verbiage or patterns that you see across different 10Ks would be super helpful! @bobmahalo
bobmahalo commentedon Apr 28, 2025
not sure if this screenshot will upload....test run.
JamesAlfonse commentedon Apr 29, 2025
Yup I see it! Thanks Bob. Can you fit them into a comment, or do a pull request to add them into the repo? It'll be easier to work with @bobmahalo
bobmahalo commentedon Apr 29, 2025
Raja dat. I will pelt with comment when I get to computer. Prolly in The morning.
bobmahalo commentedon Apr 29, 2025
this is the list I was able to cobble together(6days found the last term)Holders of record, Registered holders, Share holders of record, Stockholders of record, Stockholders of record, Registered holders of record, Registered holders of record, Share owners of record, Registered shareholders, Common shareholders of record, edit, Registered stockholder, Record owner. if this needs to be itemized differently, LMK.
JFWooten4 commentedon May 20, 2025
You guys are awesome hashing out the details and getting everyone compiled together in here. 馃挏